Grammars. Prof. Andreas Prinz. Introduction, Compilers. Examples Meta-models vs. Grammars Summary

Transcription

1 Meta-models and Grammars Prof. Andreas Prinz Introduction, Compilers Modelling & Meta-modelling Examples Meta-models vs. Grammars Summary

2 Compilers Graphical Editor Parser Text editor Static checks (OCL) Abstract Syntax Meta-model dl MOF-structure Exchange Format (XMI, ASN.1) Test case derivation Code generation Proofs Solved: many input/output p formats Graphical / Domain specific languages, many transformations Internal representation: Meta-model vs. Abstract syntax Prepared by Andreas Prinz 2

3 Importance of internal structure (html web pg) page) Representation (xml check descr.) generated code (python code) (web page model) Model (checking model) generated Model (code model) (access. checker) checking Tools transformation tool generated tool (type,consistency) t (checks compiler) (access. checker) Prepared by Andreas Prinz 3

4 Aspects of Compilers/Languages Language structure: What are the concepts? How are they related? Static semantics: additional conditions, what is allowed? Representation: How are programs written? -> graphical vs. textual Dynamic semantics: What do the programs mean? How to generate code for them? Prepared by Andreas Prinz 4

5 Aspects of a language & tools Graphical editor graphical run Representation Behaviour textual Structure transform Simulator Transformator Textual editor Exchange format Parser Access interface Repository Constraints Checker XMF Mosaic from Xactium as example tool set. Idea: The metamodel IS the tool. Prepared by Andreas Prinz 5

6 Aspects for SDL and UML Representation graphical Behaviour run informal EBNF formal ASM Structure textual formal EBNF transform Representation graphical Behaviour run informal informal Structure textual formal meta-model transform Constraints formal PC1 Constraints formal OCL SDL UML Prepared by Andreas Prinz 6

7 What is a Model? A model is an abstraction of a (part of a) system. one model describes several systems, one system can have several models simplified view of a system with respect to criteria can answer questions about the system if related to the view needs a representation, e.g. using a modelling language Models on different abstraction levels Models of the real Bits: Assembler Models of the Control Flow: Prog. Lang. Models of data storage: Database descriptions Models of access: Interface languages What is the best model of a cat? It is a cat. But it has to be the same cat! A model has aspects like a language. Prepared by Andreas Prinz 7

8 What are Meta-Models? A description of a class of models Models / high-level descriptions of the modelling language narrow view: structure of the modelling language g wider view: all important aspects of the language, i.e. structure, presentation, static ti and dynamic semantics Meta-models (languages) can have several aspects. Prepared by Andreas Prinz 8

9 Language support MDA and Eclipse HUGN Action Representation Behaviour Structure HUTN MOF QVT GEF/GMF --- Representation Behaviour Structure Xtext Xtend EMF Xpand Constraints OCL Constraints OCL MDA Eclipse (oaw) Prepared by Andreas Prinz 9

10 A meta-modelling architecture «component» MOF OMG Level Examples Grammar example OCL example M3 «component» UML Metamodel «component» UML Model «component» User Data M2 M1 3 = meta MOF EBNF MOF meta model 2 = meta UML MM Java OCL model grammar language 1 = model UML Model a program a formula 0 = real objects A run a truth instances value M0 Prepared by Andreas Prinz 10

11 Instances on several levels Class name : String isabstract:boolean M3 Property name:string * owned- Attribute Class:Class name = Class isabstract =false name : String isabstract:boolean M2 Class name : String isabstract:boolean Class:Class name = Class isabstract =false M1 Person :Property name = isabstract :Property name= name M0 :Person Prepared by Andreas Prinz 11

12 Simple sample structure (EMF) Puzzle dimension Field 1..* Column Row column row field cells Cell cellvalue l cellsb cellsr cellsc Box box Prepared by Andreas Prinz 12

13 Simple sample constraints (OCL) context Field inv uniqueicellvalues: self.cells->forall(c1,c2 : Cell c1<>c2 implies c1.icellvalue <> c2.icellvalue) context Cell inv rowfromcell: self.row -> size()=1 context Puzzle inv numberofboxes: self.elements->select(f : Field f.oclistypeof(box)) -> size()=9 Prepared by Andreas Prinz 13

14 Simple sample text syntax (TEF) syntax toplevel PuzzleTpl, ecorepath " { element CellTpl for Cell{ single for icellvalue, with INTEGER; } element RowTpl for Row{ "Row"; "("; sequence for cellsinrow, seperator ",", last false; ")"; } element PuzzleTpl for Puzzle{ "Puzzle"; "("; single for idimension, with INTEGER; ")"; "="; sequence for Elements, seperator ",",, last false; } choice FieldTpl for } } Prepared by Andreas Prinz 14

15 Simple sample graphics Sudoku Diagram rows subsets diacontents Row Container cells subsets contents Cell Rectangle Prepared by Andreas Prinz 15

16 Simple sample transformation (QVT) transformation theone (source : sudoku, target: sudoku){ top relation change1to16 { checkonly domain source sudoku:cell { icellvalue = 1 }; enforce domain target sudoku:cell { icellvalue = 16 }; } top relation change6to11 { checkonly domain source newstructure:cell { icellvalue = 6 }; enforce domain target newstructure:cell { icellvalue = 11 }; } top relation nochange { value: Integer; checkonly domain source newstructure:cell { icellvalue = value }; enforce domain target newstructure:cell { icellvalue = value }; when{ icellvalue <> 1 or icellvalue <> 6; } } } Prepared by Andreas Prinz 16

17 Simple sample execution Run(s:Sudoku) = def forall f in self.field do RunF(f) Runf(f:Field) = def choose c in self.cell with c.value=null and c.possible.size = 1 choose v in c.possible do c.value:= v choose c in self.cell with c.value<>null forall cc in self.cell do delete c.value from cc.possible Prepared by Andreas Prinz 17

18 Problem area execution Syntax Runtime Meta-model Cell RTCell Model X:Cell A: RTCell B: RTCell e.g. history Prepared by Andreas Prinz 18

19 Problem area representation There are usually several representations for the same meta-model instances. Tools and theory exist only for the case 1:1. A representation is a separate model that is related to the meta-model. Sig1: SignalDefinition signal Sig1, Sig2; signal Sig1; signal Sig2; signal Sig1; Sig2: SignalDefinition signal Sig2; Prepared by Andreas Prinz 19

20 Meta-models versus grammars Advantages of grammars Strong mathematical basis Tree-based Trees can be extended into general graphs Several advanced tools available Easily understandable Advantages of meta-models Direct representation of graphs (graphics!) Namespaces and relations between language elements (in particular for language transformations and combinations) Object-oriented definition of oo languages More problem-oriented Reuse and inheritance Tools allow direct handling of models (repositories) Structuring possible (e.g. packages) Prepared by Andreas Prinz 20

21 Grammars meta-models 1. Every symbol is represented with a class. 2. A rule with a single symbol on the rhs is represented with an association between the class representing the lhs and the rhs. 3. A rule with a composition on the rhs is represented with an association for every sub-expression. 4. A rule with an alternative on the rhs is represented with a generalization for every sub-expression. 5. A sub-expression consisting of just one symbol is represented with the symbol s class. 6. A sub-expression being a composition or an alternative is represented with a new class with new name. The composition is then handled like a rule. Prepared by Andreas Prinz 21

22 Using the transformation for SDL Joachim Fischer, Michael Piefel, Markus Scheidgen: A Metamodel for SDL-2000 in the Context of Metamodelling ULF in Proceedings of SAM2006 Introduction of abstract concepts General: namespace, namedelement, typedelement Specific: parametrizedelement, bodiedelement Introduction of relations Procedure name versus procedure definition Deletion of grammar artefacts Referencing: identifier, qualifier Names in general Superfluous structuring Prepared by Andreas Prinz 22

23 Conclusions / Summary Future languages g will be defined using meta-models. definition of good meta-models is difficult need also agreement (standard) patterns for good models needed, maybe joint concepts Meta-models / Languages have several aspects: structure, syntax, static and dynamic semantics Meta-model language definitions allow tool generation Direct access to the models Easy exchange of representation or several of them Combination of tools handling the language Description of relations between languages Meta-models are the models to be used in model- driven compiler technology. Prepared by Andreas Prinz 23